1. SUMMARY

Maternal smoking during pregnancy is a well-established and preventable risk factor for low birthweight and its sequelae of poor physical, cognitive and behavioural development, excess morbidity, and increased rates of both perinatal and adult mortality. Preliminary findings from genetic epidemiology studies suggest that the degree to which prenatal tobacco exposure depresses birthweight may be moderated by foetal and maternal genotype. The advent of affordable genomewide genotyping presents an opportunity for systematic investigation of this phenomenon. We propose to prioritize candidate polymorphisms using genomewide genotype data from three studies for which data is available on prenatal tobacco exposure and perinatal outcomes. We will use this information to prioritize approximately 50 single nucleotide polymorphisms (SNPs) to be genotyped in 3,500 mother-child dyads participating in ALSPAC. We will use these data to test hypotheses about the moderating effects of maternal and foetal genotype on the effects of prenatal tobacco exposure to reduce birthweight, shorten gestation, and alter postnatal physical, cognitive, and behavioural development.

2. BACKGROUND

Maternal smoking during pregnancy is a well-established and preventable risk factor for low birthweight (less than 2,500g) and its sequelae of poor physical, cognitive and behavioral development, excess morbidity, and increased rates of both perinatal and adult mortality (Kramer, 2003). The primary causes of low birthweight are preterm birth and intrauterine growth restriction (IUGR). The consequences of IUGR include both short-term and long-term morbidity and permanent deficits in growth and neurocognitive development (Kramer, 2003). Epidemiological studies have shown that maternal smoking is associated with both short gestation and IUGR (Kramer, 2003). Mothers who quit smoking while pregnant have longer gestations and heavier newborns than those who continue to smoke (Lumley, Oliver, Chamberlain, & Oakley, 1998). This fact is recognized in public policy: in the UK, formal smoking cessation programs are recommended as part of antenatal care to prevent low birthweight (Health Development Agency, 2004).

Preliminary findings from studies in genetic epidemiology, including my own research, suggest that the degree to which prenatal tobacco exposure depresses birthweight may be moderated by foetal and maternal genotype (e.g. Infante-Rivard, Weinberg, & Guiguet, 2006, T. S. Price, Grosser, Plomin, & Jaffee, 2008). The combination of affordable experimental methods for high-throughput genotyping and the availability of richly phenotyped birth cohorts for whom information is available on prenatal environmental exposures and perinatal outcomes provides an opportunity for systematic investigation of this phenomenon. I propose to study the issue using a two-phase experimental design. Phase 1 will address the issue broadly, in three large cohorts, by studying the impact on intrauterine growth of all common genetic variation. Phase 2 will address the issue deeply, in an independent sample, by looking at the downstream effects of selected polymorphisms on long-term features of postnatal development.

3. METHODS

3.1 Overview

Phase 1. Genomewide association studies to detect interactions between maternal and child genotype and maternal smoking during pregnancy on intrauterine growth. I have negotiated access to three studies with genomewide genotype data and information on prenatal environmental exposures and perinatal outcomes. Two of these studies are birth cohorts with genomewide genotype information on the children: the Twins Early Development Study (TEDS, PI R. Plomin; Trouton, Spinath, & Plomin, 2002) and the 1966 North Finnish Birth Cohort (NFBC, PI: M.-R. Jarvelin; Rantakallio, 1969). The third has genotype data from parent-child trios and siblings (IMAGE, PI: P. Asherson; Kuntsi, Neale, Chen, Faraone, & Asherson, 2006). In total, these samples comprise more than 9,000 families for whom information will be available on genomewide genotype, environmental risk exposure and perinatal outcomes. Hypotheses of genotype*environment interaction and haplotype*environment interaction will be tested genomewide in each study population. Untyped polymorphisms will be imputed using standard techniques (Marchini, Howie, Myers, McVean, & Donnelly, 2007) to ensure that the genotype set is consistent across studies and facilitate meta-analysis.

Phase 2. Replication and extension. Approximately 50 single nucleotide polymorphisms (SNPs) prioritized in the initial stage - located in the 5 genomic regions providing the best evidence of genotype-environment interaction - will be typed in an independent sample of approximately 3,500 mother-child dyads participating in the Avon Longitudinal Study of Parents and Children (ALSPAC, PI: G. Davey-Smith; Golding, Pembrey, & Jones, 2001). These data will be used to test hypotheses of genotype-environment interaction and haplotype-environment interaction on intrauterine growth and on postnatal phenotypes associated with maternal smoking during pregnancy including Attention-Deficit Hyperactivity Disorder (ADHD), cognitive ability, and height. The availability of prospective information on environmental exposure in this sample will allow hypotheses to be tested about dose-related effects and the impact of timing for mothers who quit smoking during pregnancy.

3.2 Data collection.

We propose to genotype approximately 3,500 mothers and their children participating in the ALSPAC study for approximately 50 single nucleotide polymorphisms prioritized in the first phase of the study. We will choose the 1,500 or so families in which the mother reported smoking throughout the pregnancy, an equal number of families in which the mother reported not smoking at any time during the pregnancy, plus the 500 or so families in which the mother reported smoking early in the pregnancy and subsequently quitting. The intention is to select the five regions providing the best evidence of genotype-environment interaction in phase 1, and fine map the immediate regions of linkage disequilibrium (as defined using resequencing data for Caucasian populations) in ALSPAC so as to be able to test hypotheses of haplotype-environment interaction in phase 2. Although haplotype block length is enormously variable, we anticipate an average of 10 non-redundant SNPs per region will suffice based on estimates of haplotype block length in Caucasian populations (Li & Chen, 2008).

3.3 Existing data required

Concept

Specific Measure

Person

Source

Time Point(s)

Demographic variables (age, sex, ethnicity, marital status, family structure, SES, education, employment, income etc.)

Family

Questionnaire

Antenatal

Pregnancy health variables (nulliparity, pre-eclampsia, IUGR, gestational diabetes etc.)

Mother

Questionnaire, medical records

Antenatal

Parental anthropometrics (height weight)

Mother, Father

Questionnaire

Antenatal

Pregnancy exposure variables (tobacco, alcohol and drug use; chemical exposure; diet; nutrient supplementation; stress; life events; partner cruelty; lack of social support)

Mother

Questionnaire

Antenatal

Birth/delivery variables (weight/length, placental weight, gestational age, Caesarian), perinatal health

Child

Questionnaire, medical records

Birth/perinatal period

Childhood head injury

Child

Questionnaire

Birth - 4 years

Childhood temperament

Carey

Child

Questionnaire

6-24 months

Childhood behaviour

SDQ

Child

Questionnaire

42 - 157 months

ADHD

Child

Questionnaire

166 months

Antisocial behaviour

Child

Questionnaire

169 months - 198 months

Psychotic symptoms

Child

Questionnaire

140 months - 198 months

Adolescent substance use (tobacco, alcohol, drugs)

Child

Questionnaire

157 months - 198 months

Language development

McCarthy

Child

Questionnaire

24 months

Cognitive ability

WISC

Child

Clinic test

8-10 years

Scholastic achievement

Child

Questionnaire

166 months

Anthropometrics (Height, weight)

Child

Questionnaire

Birth - 157 months

Parental antisocial behaviour (antisocial behaviour as children or adults; contact with police; criminal convictions)

Mother, father

Questionnaire

Antenatal - 145 months

Childhood postnatal experiences (maternal depression and anxiety, parental discipline)

Mother

Questionnaire

Birth - 145 months

Parental postnatal substance use (tobacco, alcohol, drugs)

Mother, father

Questionnaire

Birth - 145 months

3.4 Data Analysis.

Phase 1. Missing genotype data (including polymorphisms genotyped in at least one but less than three of the datasets) will be imputed using published methods (Marchini et al., 2007). Hypotheses about effects on birthweight (corrected for gestational age) will be tested using linear models incorporating terms for prenatal tobacco exposure, foetal genotype, interaction between foetal genotype and prenatal tobacco exposure plus relevant covariates (including significant principal components of the genetic data to guard against spurious associations due to population admixture (A. L. Price et al., 2006), demographics, pregnancy and antenatal health variables, and parental physical and behavioural characteristics). Meta-analysis will be performed to synthesize the results from the three studies. SNPs to be genotyped in phase 2 will be chosen based on the regions that show the greatest evidence of association in phase 1. Where significant results are found for foetal genotype, post hoc analyses of the influence of maternal genotype and parent-of-origin effects will be tested in the IMAGE sample.

Phase 2. Hypotheses about effects on birthweight, head circumference, and crownheel length (all corrected for gestational age) will be tested using linear models incorporating terms for prenatal tobacco exposure, maternal genotype, foetal genotype, interaction between maternal genotype and prenatal tobacco exposure, interaction between foetal genotype and prenatal tobacco exposure, and the three-way interaction between foetal and maternal genotype and prenatal tobacco exposure, plus relevant covariates (including demographics, pregnancy and antenatal health variables, and parental physical and behavioural characteristics). Analyses will be stratified by ethnicity to guard against spurious associations due to population admixture. Should statistically significant results be found, post hoc hypotheses about trajectories of postnatal physical, behavioural, and cognitive development will be tested using growth curve models and growth mixture models; in addition, mediation analyses will be conducted to test whether postnatal development is conditionally independent of genotype and prenatal tobacco exposure after controlling for any effects on birthweight and gestational age. Hypotheses about the effects of prenatal tobacco exposure will test for heterogeneity of the effects with respect to mode of exposure (maternal smoking/maternal exposure to second hand smoke), dose, and timing of smoking cessation. A parallel set of analyses will be conducted using paternal tobacco exposure as the environment of interest in order to validate the inference of intrauterine effects. Analyses will be conducted to assess the possible influence of attrition in the sample on the outcomes of interest. If necessary, informative missingness will be explicitly modelled.

3.5 Statistical Power

Phase 1. Based on the most accurate information available, we anticipate that information on genomewide genotype, prenatal environmental exposures, and perinatal outcomes will be available for children from 4,000 families enrolled in TEDS, 4,763 families participating in NFBC, and 304 families from IMAGE. We expect that the numbers of children born to mothers who reported smoking during pregnancy in these three samples will be, respectively, 568, 700, and 63. For the purposes of the power analysis let us assume a total population of N = 9,067, exposure prevalence of 14.7%, population SD of birthweight = 400g, marginal effect of maternal smoking -200g, and no marginal effect of genotype. We will use alpha = 10-6 so that if we test 2-sided hypotheses of GxE in relation to 10-6 SNPs we expect on average 1 false positive by chance. Under these assumptions we will have 80% power to detect GxE accounting for 0.356% of the variance in birthweight and 50% power to detect GxE accounting for 0.261% of the variance.

Under an additive model of inheritance, 0.261%-0.356% of the variance corresponds to GxE accounting for a difference of 102-119 g between groups differing by an allele count of 1 and with different smoking status, assuming a minor allele frequency of 20%, or 136-159 g assuming a minor allele frequency of 10%. Under a dominant model of inheritance, 0.380%-0.521% of the variance corresponds to GxE accounting for a difference of 120-141 g between groups differing in whether or not they carry a risk allele and with different smoking status, assuming a minor allele frequency of 20%, or 147-172 g assuming a minor allele frequency of 10%.In other words, for common SNPs there is good power to detect a GxE whose coefficient is similar in magnitude to the marginal environmental effect. Effect sizes of this magnitude are by no means implausible, as recent candidate gene studies have demonstrated (e.g. Sasaki et al., 2008, T. S. Price et al., 2008), but there are likely to be relatively few variants in the genome that could account for interactive effects of this size. It is for this reason that we intend to follow up only a handful of the best hits.

Phase 2. Assuming that genotype data is available for 3,000 individuals, with 50% exposure to prenatal smoking throughout pregnancy, a population SD for birthweight of 400g, a marginal effect of maternal smoking of -200g, and using 2-tailed hypotheses for each of the three phenotypes of birthweight, head circumference, and crownheel length and a conservative significance threshold of alpha = 3.3e-4 (0.05 after Bonferroni correction for 50 x3 = 150 hypothesis tests), we estimate that for the effect sizes giving 50% power in the phase 1 we expect 69% power to replicate in phase 2, and for the kinds of effect size giving 80% power in phase 1 we expect at least 88% power to replicate in phase 2.

3.6 Work already completed

We recently investigated the relations between maternal smoking, foetal genotype, and foetal growth in a dizygotic twin pairs participating in TEDS (T. S. Price et al., 2008). Maternal smoking retarded growth by 118g in twins born to mothers who reported smoking less than 10 cigarettes per day and by 185g in twins born to heavier smokers, allowing for the effects of twin sex, birth order, gestational age, and maternal and familial characteristics. We selected 497 twin pairs, whose mothers smoked, for a molecular genetic study. In this subsample, a functional SNP in the NQO1 gene (Pro187Ser; rs1800566) was significantly associated with foetal growth within families assuming a dose-related model of effects (p=0.0028). These results provide the first demonstration that foetal genotype for a xenobiotic metabolizing enzyme influences intrauterine growth under conditions of smoke exposure independently of maternal genotype.

4. REFERENCES

Golding, J., Pembrey, M., & Jones, R. (2001). ALSPAC--the Avon Longitudinal Study of Parents and Children. I. Study methodology. Paediatr Perinat Epidemiol, 15(1), 74-87.

Health Development Agency. (2004). The evidence of effectiveness of public health interventions - and the implications.

Infante-Rivard, C., Weinberg, C. R., & Guiguet, M. (2006). Xenobiotic-metabolizing genes and small-for-gestational-age births - Interaction with maternal smoking. Epidemiology, 17(1), 38-46.

Kramer, M. S. (2003). The epidemiology of adverse pregnancy outcomes: An overview. Journal of Nutrition, 133(5), 1592S-1596S.

Kuntsi, J., Neale, B. M., Chen, W., Faraone, S. V., & Asherson, P. (2006). The IMAGE project: methodological issues for the molecular genetic analysis of ADHD. Behav Brain Funct, 2, 27-27.

Li, J., & Chen, Y. (2008). Generating samples for association studies based on HapMap data. BMC Bioinformatics, 9, 44-44.

Lumley, J., Oliver, S., Chamberlain, C., & Oakley, L. (1998). Interventions for promoting smoking cessation during pregnancy. Cochrane Database of Systematic Reviews(3), Art. No.: CD001055. DOI: 001010.001002/14651858.CD14001055.pub14651852.

Marchini, J., Howie, B., Myers, S., McVean, G., & Donnelly, P. (2007). A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet, 39(7), 906-913.

Price, A. L., Patterson, N. J., Plenge, R. M., Weinblatt, M. E., Shadick, N. A., & Reich, D. (2006). Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet, 38(8), 904-909.

Price, T. S., Grosser, T., Plomin, R., & Jaffee, S. R. (2008, Nov 11-15). Fetal genotype for the xenobiotic metabolizing enzyme NQO1 influences intrauterine growth among infants whose mothers smoked during pregnancy. Paper presented at the American Society for Human Genetics, Philadelphia, PA.

Rantakallio, P. (1969). Groups at risk in low birth weight infants and perinatal mortality. Acta Paediatr Scand, 193.

Sasaki, S., Sata, F., Katoh, S., Saijo, Y., Nakajima, S., Washino, N., et al. (2008). Adverse birth outcomes associated with maternal smoking and polymorphisms in the N-nitrosamine-metabolizing enzyme genes NQO1 and CYP2E1. American Journal of Epidemiology, 167(6), 6.

Trouton, A., Spinath, F. M., & Plomin, R. (2002). Twins Early Development Study (TEDS): A multivariate, longitudinal genetic investigation of language, cognition and behavior problems in childhood. Twin Research, 5(5), 444-448.